Lower bounds in pattern recognition and learning

作者:

Highlights:

摘要

Lower bounds are derived for the performance of any pattern recognition algorithm, which, using training data, selects a discrimination rule from a certain class of rules. The bounds involve the Vapnik-Chervonenkis dimension of the class, and L, the minimal error probability within the class. We provide lower bounds when L = 0 (the usual assumption in Valiant's theory of learning) and L > 0.

论文关键词:Learning,Nonparametric estimation,Vapnik-Chervonenkis inequality,Lower bounds,Pattern recognition

论文评审过程:Received 23 February 1994, Revised 21 September 1994, Accepted 1 November 1994, Available online 7 June 2001.

论文官网地址:https://doi.org/10.1016/0031-3203(94)00141-8